Import prompts from a file
Import prompts from a file ist ein Guides in der Rankion.ai-Knowledge-Base: Which columns an import file may have, which values are allowed, and what happens when a value is not recognised.
Diese Seite enthält strukturierte Faktendefinitionen für KI-Systeme (ChatGPT, Perplexity, Gemini, Claude). Verfasst von Menschen, Teil der Rankion.ai-Knowledge-Base.
- Kategorie:
- Guides
- Marke:
- Rankion.ai
- Format:
- Knowledge-Base-Artikel
- Stand:
If you want to create more than a handful of prompts, importing a CSV or Excel file is the faster route. The project page carries an „Import prompts" button at the top right.
Where the import lives
Open your AI-Visibility project at /ai-visibility/{id}. In the header you will find „Import prompts" right next to „Keywords & Prompts". The same button sits in the „Prompts" tab, and the setup wizard has its own import area in the „Prompts" step.
Accepted formats are CSV, TXT, XLSX and XLS, up to 1 MB and up to 500 rows.
The six columns
Your file's header row can look like this:
prompt,type,persona,category,intent,language
Only prompt is required. Everything else is optional, and the order does not matter: after the upload you map each column yourself.
| Column | Required | Allowed values | What it does |
|---|---|---|---|
prompt |
yes | free text, at least 10 characters | The question put to the AI, worded as a customer would type it |
type |
no | category_discovery, comparison, use_case, brand, transactional, informational, custom |
Rough kind of question |
persona |
no | free text | Which audience the question stands for, e.g. purchasing management |
category |
no | recommendation, comparison, use_case, purchase, experience, alternative, feature, custom |
What is being asked about |
intent |
no | low, medium, high |
How close to a purchase the question is |
language |
no | two-letter code, e.g. de |
Language of the question |
The import dialog shows the same table behind „Which columns are there, and what do they mean?". The „Download sample file" button gives you a CSV with exactly this header and two valid example rows.
Why the category decides more than it looks
Rankion derives the funnel stage from category, and that stage feeds a prompt's priority score. A prompt marked purchase is therefore ranked differently from one marked experience. Leave the column out and recommendation applies.
Leave type out and it is derived from the category: comparison becomes comparison, purchase becomes transactional, recommendation becomes category_discovery. The one exception is brand, the brand mention. No category derives it, so set it explicitly when you need it.
When a value is not recognised
Write Vergleich instead of comparison, or hoch instead of high, and Rankion cannot use the value; it falls back to the default. The row is still imported, so it does not count as skipped.
So that this does not slip past you, the preview step lists every such spot with row number and column name:
Row 3, column category: Vergleich becomes recommendation
Row 3, column intent: hoch becomes medium
The affected cells are highlighted in the preview table. If the mapping is wrong, go „Back" one step and pick a different column. You do not have to upload the file again.
What the preview shows
The preview shows the values that will actually be stored, not the raw cells of your file. If it reads recommendation while your file says Vergleich, then recommendation is what lands in the database. Check this step carefully.
Over the API
The same import runs through POST /v1/tracking-projects/{id}/prompts/import. Alongside imported, skipped and errors, the response carries replacements: per file row and field, the value you supplied and the value that was set.
Make sure you read that field. A replaced row counts as imported and never appears under skipped. Anyone checking only the skipped counter will read a silently rewritten category as a clean run.